Meta scrapes 70.4% of all training crawler volume

So Meta is: - 31.3x OpenAI training volume - 29.2x Anthropic training volume - 70.4% of all training crawler volume across the dataset n=84 DataFast websites with at least 1,000 training crawls in the last 7 days. They scrape with the user agent `meta-externalagent` used for "training foundation AI models or improving products ".

Marc Lou@marclou · Jul 5, 2026Meta is scraping like crazy. AI training crawls last 3 days on TrustMRR: - OpenAI: 409 - Anthropic: 721 - Meta: 67,390